人工生命进化的可能规律
Possible Laws for Artificial Life Evolution
摘要:
受最近一篇关于人工生命中的开放问题的文章的启发,在这里我假设三个定律,它们形成了一个数学框架来描述人工生命进化动力学。它们是基于人口动态的连续近似。在这个公式中需要四个动态元素:优势矩阵,横向矩阵,适应度函数,和随机驱动器。第一定律指出,在没有随机驱动的情况下,人工生命总是寻找一个局部适应吸引子并停留在那里。为讨论一般演化动力学提供了参考点。第二定律被明确地表达为具有所有四个动力学元素的随机微分方程的唯一形式。第三定律定义了描述的重点层次与其较低和较高层次之间的关系,也定义了确定性驱动和随机驱动的二分法。这些法律为讨论当前的几个问题,如紧急状态和稳定,提供了一个协调一致的框架。特别强调了两个量:作为选择标准的适应度函数和作为创造力源泉的随机性。从统计物理学的观点来看,这三个定律似乎几乎是不言自明的。然而,他们相当于一个最传统的方法进化动力学是第一次由本作者,尽他所知。本文还讨论了该方法在人工生命进化研究中的计算优势。
英文摘要:
Motivated by a recent article on open problems in artificial life, here I postulate three laws which form a mathematical framework to describe artificial life evolutionary dynamics. They are based on a continuous approximation of population dynamics. Four dynamical elements are required in this formulation: ascendant matrix, transverse matrix, fitness function, and the stochastic drive. The first law states that in the absence of stochastic drive the artificial life always seeks for a local fitness attractor and stay there. It gives the reference point to discuss the general evolutionary dynamics. The second law is explicitly expressed in a unique form of stochastic differential equation with all four dynamical elements. The third law defines the relationship between the focused level of description to its lower and higher ones, and also defines the dichotomy of deterministic and stochastic drives. These laws provide a coherence framework to discuss several current problems, such as emergency and stability. In particular, two quantities are emphasized: the fitness function as the standard for selection and the stochasticity as the source of creativity. Those three laws may appear almost self-evident from a statistical physics point of view. However, their equivalent to a most conventional approach for evolutionary dynamics is shown for the first time by the present author, to the best of his knowledge. The computational advantage of the present formulation in the study of artificial life evolution is also discussed.
机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。